Major Depressive Disorder (MDD) is associated with structural, functional, and neurochemical alterations in key interrelated brain circuits involved in emotion, reward, and executive functioning. Current models of its etiology, including genetic expression, gene environment interactions, the monoamine hypothesis, and neurogenesis guided our choice of biomarkers. We propose to use biomarkers from several levels of organization that address one or more of these models and examine their ability to predict treatment remission. At the genetic level, we will examine epigenetic measures and the transcriptome. At the molecular level, the utility of measures of 5HT1a neuroreceptor binding using Position Emission Tomography and proteomics will be investigated. At the anatomical level, we will examine white matter tract integrity and regional decreases in cortical thickness. Functional assessments include electroencephalography, loudness dependent auditory evoked potentials, and neurocognitive performance. Clinical features will be studied as well, e.g. presence of anxious depression, family history of depression, and others. While receiving supportive clinical management, 300 patients will be observed medication free for 3 weeks, to diminish the influence of placebo response and minimize effects on biosignature assays. Those still meeting criteria after the 3 weeks will receive all aforementioned assessments. Patients then will be randomized in a doublemasked fashion to bupropion or escitalopram, two of the most commonly prescribed treatments for depression, with putatively distinct mechanisms of action. Treatment will be for 12-14 weeks. Treatment outcome will be remission, measures of symptomatic improvement, and assessment of adverse events. Non-remitters will be crossed over. Outcomes will be measured with both traditional and contemporary clinical assessments. Patients will be followed for 6 months after randomization to assess maintenance of response and remission. We will also use a comprehensive analysis algorithm, using novel statistical techniques for high dimensional data to develop an optimal predictive model of treatment outcome that includes all data recorded from all modalities. The statistical team will develop new strategies to address the complex data set to be generated by this study. The resulting optimized algorithm for predicting remission can serve as the basis for a new study intended to validate this tool for personalized treatment of depression. Data and biological materials collected in this project would become part of a repository, open to qualified individuals for additional analysis.